International Organization Foundation

International Organization Foundation
The Rational Design of International Institutions Author(s): Barbara Koremenos, Charles Lipson, Duncan Snidal Source: International Organization, Vol. 55, No. 4, The Rational Design of International Institutions (Autumn, 2001), pp. 761-799 Published by: The MIT Press Stable URL: http://www.jstor.org/stable/3078615 Accessed: 03/11/2009 10:56
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=cup and http://www.jstor.org/action/showPublisher?publisherCode=mitpress. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Cambridge University Press and International Organization Foundation are collaborating with JSTOR to digitize, preserve and extend access to International Organization.
http://www.jstor.org
The Rational Design of International Institutions

BarbaraKoremenos, Charles Lipson, and Duncan Snidal
relations.This institutionsare centralfeaturesof moderninternational International financial restructuring,and even national is true of trade, internationaldebt and security, once the exclusive realm of pure state action. It was certainly true of the two major military engagements of the 1990s, the wars in Kosovo and the Persian Gulf. As internationalinstitutions have gained prominence in the political landscape, they have increasingly become prominent topics for study. The sharpest debate among researchershas been theoretical:Do internationalinstitutionsreally matter?Missing from this debate is a sustainedinquiry into how these institutions actually work. We shift the focus by posing researchablequestions about how they operate and how they relate to the problems states face. We begin with a simple observation:majorinstitutionsare organizedin radically different ways. Some are global, essentially open to all states; others are regional, with restricted memberships. Some institutions give each state an equal vote, Instiwhereas others have weighted voting and sometimes requiresupermajorities. authorities and significant operating tutions may have relatively strong central responsibilitiesor be little more thanforumsfor consultation.Some arrangementsAs this project came to fruition, we received valuable input from many sources. We thank Kenneth Abbott, George Downs, James Fearon, Phillip Genschel, Charles Glaser, Lloyd Gruber,Miles Kahler, Robert Keohane, Dan Lindley, Lisa Martin, Ken Oye, Beth Yarbrough,Alexander Thompson, Mark Zacher, and especially Brian Portnoy, who participatedin one or more of the conferences leading up to David Laitin,Joni Harlan,and JamaAdams this volume. Jeffrey Smith, Ryan Peirce, MarcTrachtenberg, provided other valuable comments, as did the participantsat the Program on InternationalPolitics, Economics, and Security (PIPES), University of Chicago, where this project began. Students who participatedin Barbara Koremenos' undergraduateseminar at UCLA, "InternationalCooperation," for providedvaluable feedback. We also thankthe contributors their efforts, not only on their individual articles but also on the design of the project as a whole. James Morrow, Ronald Mitchell, Peter Rosendorff,Robert Pahre, and especially Andrew Kydd contributedgreatly to the project. We received invaluablecriticism, prodding,and supportfrom two anonymousreviewers, from the editors of IO, and from Lynne Bush. We thank the University of Chicago's Council on Advanced Studies on Peace and International Cooperationfor funding supportand the HarrisGraduateSchool of Public Policy Studies for hosting the Rational Design conferences. Finally, we thank Loch Macdonald,BarbaraKoremenos' neurosurgeon,who was there when we needed him. InternationalOrganization55, 4, Autumn 2001, pp. 761-799 ? 2001 by The IO Foundationand the MassachusettsInstituteof Technology
762
International Organization
for example, most bilateraltreaties-have no formal organizationalstructure; these are plentiful because states have a strikingtendency to codify their relationshipsin formal, legal arrangements.' Why do these differencesexist? Do they really matter,both for membersand for international themselves politics more generally?Do they affect what the institutions can do? We focus on these largequestionsof institutional design. Ourbasic presumption, groundedin the broad traditionof rational-choice analysis, is that states use internationalinstitutionsto further their own goals, and they design institutions controversial. accordingly.This might seem obvious,but it is surprisingly One critiquecomes from constructivists,who arguethat international institutions play a vital, independentrole in spreadingglobal norms. We agree that normative discourse is an importantaspect of institutionallife (though surely not the whole of it) and that norms are contested within, and sometimes propagatedby, international institutions. But it is misleading to think of internationalinstitutions solely as outside forces or exogenous actors. They are the self-conscious creation of states (and, to a lesser extent, of interest groups and corporations). The realist critiqueis exactly the opposite. For them, international institutionsare little more than ciphers for state power. This exaggeratesan importantpoint. States rarely allow internationalinstitutions to become significant autonomous actors. Nonetheless, institutions are considerably more than empty vessels. States spend significant amounts of time and effort constructinginstitutions precisely because they can advance or impede state goals in the internationaleconomy, the environment, and national security. States fight over institutionaldesign because it affects outcomes. Moreover,the institutionsthey createcannotbe changed swiftly or easily to conform to changing configurationsof international power. Japanand Germany modest roles in the UN today because they have been unable to reverse the play decision made in 1944-45 to exclude them from the Security Council. Institutions rarelyadaptimmediatelyto states' growing (or ebbing) power. For this reason, and because institutionsmatter,states pay careful attentionto institutionaldesign. Ourmain goal is to offer a systematicaccountof the wide rangeof design features that characterizeinternationalinstitutions. We explore-theoretically and empirically-the implications of our basic presumptionthat states construct and shape institutions to advance their goals. The most direct implication is that design differences are not random.They are the result of rational,purposive interactions among states and other internationalactors to solve specific problems. We define internationalinstitutionsas explicit arrangements,negotiated among internationalactors, thatprescribe, proscribe, and/or authorizebehavior.2Explicit arrangementsare public, at least among the parties themselves. According to our definition, they are also the fruits of agreement. We exclude tacit bargains and implicit guidelines, however importantthey are as general forms of cooperation.
1. See Abbott et al. 2000; and Koremenos2000. 2. For related definitions of internationalinstitutions,see Keohane 1984; and Young 1994.
The Rational Design of International Institutions 763
Institutions may require or prohibit certain behavior or simply permit it. The arrangementsthemselves may be entirely new, or they may build on less formal arrangementsthat have evolved over time and are then codified and changed by negotiation. The 1961 Vienna Law on Treaties is a good example. negotiatorsare typically states, this is not partof Although in most arrangements our definition;it is an empirical observationthat may vary across issues and over time. In fact, nonstate actors participatewith increasing frequency in institutional design. Multinationalfirms, nongovernmentalorganizations(NGOs), and intergovernmentalorganizationshave all shaped international institutions,solely especially those dealing with the world economy, the environment,and human rights. Thus our definition of internationalinstitutions is relatively broad. It includes formal organizationslike the World Health Organizationand InternationalLabor like "diplomatic Organization,as well as well-defined (and explicit) arrangements immunity" that have no formal bureaucracyor enforcement mechanisms but are fundamentalto the conduct of internationalaffairs. With this definitionin mind, we can begin to explore how institutionsvary and, later, how that variationmay be the productof rationaldesign considerations.Our work emphasizes five key dimensions within which institutionsmay vary:
Membership rules (MEMBERSHIP)
Scope of issues covered (SCOPE)

Centralization of tasks (CENTRALIZATION) Rules for controlling the institution(CONTROL) Flexibility of arrangements (FLEXIBILITY)
These are certainlynot the only significantinstitutionaldimensions,but they have several advantages for our research. First, they are all substantively important. Negotiators typically focus on them, and so do analysts who study institutions. Second, they can be measured, allowing us to compare them within and across institutions, institutionsover time. Third,they apply to the full arrayof international from the most formal to the least bureaucratic. We locate our analysis in the rational regime tradition. We do not present a literature review but rather build on earlier work to develop the underlying parametersof this researchproject. We also do not counterpose"duelingperspectives" (realism versus institutionalism or rationalism versus constructivism, for example). Instead,we investigate the rationaldesign approachon its own terms by developing a set of theoretically based conjectures, which are then evaluated empirically in the studies in this special issue of InternationalOrganization.Our view is thatrationaldesign can explain much aboutinstitutions,but not everything.3
institutionsand propose an agenda focused 3. Martinand Simmons assess past work on international on explaining causal mechanisms and institutionaleffects. Martinand Simmons 1998. Their framework complements ours and shows how rationalchoice can address other importantempirical questions.
764
From Cooperation Theory to Rational Design The postwar study of internationalinstitutions is coming full circle, but with a theoreticaltwist. The early literaturefocused on the operationaldetails of international organizations. With the notable exception of neofunctionalist integration theory,it was heavily descriptive,4neithertheorizinginstitutionsnor clarifyingtheir relationshipsto wider issues of internationalrelations. By the 1980s the literature had turned sharply toward theory under the broad rubric of "regimes."5Within regime theory, one important strand built on rational, game-theoretic analysis, especially the idea that the "shadowof the future"can support"cooperationunder anarchy."6 The study of regimes favored theoretical questions and moved the research Likewise, the agenda away from analyzing specific institutional arrangements.7 tools of game theory were directedmainly at generaltheoreticalquestions,focusing on cooperation,not institutions,as the dependentvariable.The overridingquestion became "How could states and other international actors produce cooperative outcomes by their own, self-interestedchoices?" Indirectly,however, this work laid the foundationfor a renewed explorationof institutions,this time as partof a wider theory of international cooperation.In focusing on how self-interestedstates could cooperate,it was logical to ask what role institutionscould play. Institutionscould be reconceptualizedand theorized as arrangementsthat make cooperation more feasible and durable,at least in some circumstances. Our goal is to close the circle that began with descriptive studies by explaining majorinstitutionalfeaturesin a theoreticallyinformedway. We firstrelax some key assumptionsof cooperationtheory and then bring in institutionsdirectly by incorporatinginsights from game theory and institutionalanalysis. In doing so, we pay particularattentionto the logic of their development.
Extending Cooperation Theory The cooperation literatureis premised on the "Folk theorem,"which shows that cooperation is possible in repeated games.8 This result has a strong theoretical foundationand can be applied empiricallyto a wide range of contemporaryissues. The density of contemporaryinternationalinterdependencecreates repeated inter-
4. The early issues of InternationalOrganization,for example, focused on describingnewly formed organizationsand publicizing their rules and votes. 5. Krasner1983. 6. See Oye 1986; and Axelrod 1984. 7. Key works are Stephen Krasner's edited volume International Regimes (1983) and Robert Keohane'sAfter Hegemony(1984). An excellent early overview is Haggardand Simmons 1987. Several commentators have noted that the field has had less and less to say about formal international organizations.See Rochester 1986; and Abbott and Snidal 1998. 8. See Friedman 1971; and Fudenbergand Maskin 1986.
action that makes cooperation feasible.9 In brief, the possibility of cooperation is present in most moder internationalissues. If cooperationis within reach, why it is not always grasped?To answer that, we of must go beyond any simple, optimisticinterpretation the Folk theorem.Although we assume that the general conditions of internationalinterdependenceare propitious, individual issues have featuresthat make achieving and maintainingcooperation more problematic. Moreover, the standardFolk theorem conclusion needs careful refinement when applied to more realistic situations, where competing equilibria are in play, many actors are involved, and uncertaintyis high. Multiple equilibriaare a major obstacle to cooperationthat was downplayedby the early emphasis on 2 X 2 games. Although these simple games, especially Prisoners' Dilemma, did much to clarify our understandingof enforcementproblems, their very simplicity could be misleading. In a simple 2 x 2 Prisoners' Dilemma, there is only one point of mutual cooperation, the unattainablePareto optimumwhere both sides choose to cooperateratherthandefect. In practice,states have a wide range of choices and many possible cooperativeoutcomes, often with different distributionalconsequences. If actors prefer different outcomes, the range of possibilities creates bargaining problems. Which cooperative outcome should they choose? How, in other words, should they share any mutual gains from cooperation?These distributionalquestions do not arise in simple 2 X 2 Prisoners' Dilemma games, though they were discussed in some early work contrasting Prisoners' Dilemma and Coordination games.10Recent work by Stephen Krasner,James Morrow,and James Fearongoes further, showing how distributional differences can undermine cooperation in significant ways. Hence, distributionproblems merit at least as much attentionas enforcementproblems, which we know hamperinternationalcooperation.1 Large numbers also complicate cooperation.Kenneth Oye addressesthe collective-action problem primarilyby showing how interactionsamong large numbers can be decomposed into simple bilateral interactions.12Some issues, however, cannot be decomposed this way for technical reasons; others should not be decomposed because successful cooperationrequiresjoint action by all (as in the provision of public goods). Large numbersraise questions about how to share both the costs and benefitsof cooperation,especially when some actorsare richer,bigger, or more powerful than others. Uncertaintyis a frequentobstacle to cooperation,as is "noise,"the difficulty of observing others' actions clearly.13States are naturallyreluctantto disclose vital
9. Notable exceptions are crises where immediate incentives overwhelm longer-termconsiderations. We set such situationsaside. 10. See Snidal 1985; and Stein 1983. 11. See Krasner1991; Morrow 1994c; and Fearon 1998. 12. See Oye 1986; and Lipson 1986 for an application. 13. This point was foreshadowedby Downs, Rocke, and Siverson in their analysis of armsraces, and by Downs and Rocke in their game-theoreticanalysis of the limits to cooperation.See Downs, Rocke, and Siverson 1986; and Downs and Rocke 1990.
766
informationthat could make them more vulnerable. Reducing uncertaintyamong participantsis a major function of institutions.14 Taken together, these factors-distribution, enforcement, large numbers, and uncertainty-suggest that cooperationcan be very brittlein the real world. As these factorsvary, the prospectsfor cooperationcan shift dramatically, makingit far more difficult to manage international cooperationthan earlier,simplifiedtheories would predict.
Bringing in Institutions In broad international relations (IR) theories institutionsplay only a modest role. It is, after all, cooperation under anarchy. The primary reason for emphasizing anarchyis to rule out centralizedenforcement,but there is little considerationof the other roles institutionsmight play. In fact, institutionsoften help resolve problems of decentralizedcooperation. IR theoristshave begun to addressproblemsof cooperationin more complex and realistic settings, where there may be noise and large numbers.15 It is generally recognized that institutionsmay make cooperationmore likely,16 and the compliance literature has begun to analyze empirically how regime design promotes effective cooperation.17 So far, however, this has not developed into a more general theoreticalanalysis of specific institutionalarrangements. Our work departs significantly from the earlier cooperation literature.Because decentralizedcooperation(supported the Folk theorem)is difficultto achieve and by often brittle, states devise institutions to promote cooperation and make it more resilient. But the form these institutions take varies widely. Often the necessary institutions are fairly minimal and simply reinforce the underlying conditions for cooperation, perhaps providing the information necessary for bilateral bargains. Other times, more complex problems may requirea larger institutionalrole-such as when an issue involves actors with very different resources and information. Under these circumstances, institutions can play a major role in facilitating cooperation. We argue that many institutional arrangementsare best understood through "rationaldesign"among multipleparticipants. This rationalityis forwardlooking as states use diplomacy and conferences to select institutionalfeaturesto furthertheir individual and collective goals, both by creating new institutions and modifying existing ones. Even trial-and-error experimentscan be rationaland forwardlooking in this way. Although we do not arguethat all institutionalchange is the productof conscious design, we do consider it the overridingmechanism guiding the devel14. 15. 16. 17. See Keohane 1984; and Morrow 1994c. On noise, see Downs and Rocke 1990. On large numbers,see Pahre 1994. See Keohane 1984; and Axelrod and Keohane 1986. See Chayes and Chayes 1995; and Mitchell 1994.
Moreover, though our primarypurpose is to opment of internationalinstitutions.18 foundation explain institutionaldesign, our approachalso provides an appropriate for prescribingpolicy and evaluating existing institutions.19 Our argumentthat institutionaldesign is deliberate is reflected in the difficult process of creating an international institution. The evolution of the General Agreementon Tariffs and Trade(GATT) into the WorldTradeOrganization(WTO) involved extensive rounds of negotiation. The Law of the Sea Treaty was the culmination of protracteddebate, including the sharply contested decision not to have strongercentralizedinstitutions.The same process is seen in the development of the UN charter, which involved extensive planning and bargaining and was designed to achieve critical goals amidstgreatuncertainty.Moreover,its design has been modified over the years as new members have been admitted, the Security Council has changed, and specialized agencies have been created.Continuingcalls for change remindus that most institutionsevolve as memberslearn, new problems arise, and internationalstructuresshift. But institutional evolution still involves deliberatechoices made in response to changing conditions. Institutionaldevelopment frequently depends on prior outcomes ("path dependence") and evolutionaryforces. As institutionsevolve, rationaldesign choices can arise in two ways. First, participantsmay modify institutionsin stages, by making purposeful decisions as new circumstancesarise, by imitating features from other institutionsthat work well in similar settings, or by designing explicit institutionsto strengthentacit cooperation. Second, institutions may evolve as states (and other internationalactors) select among them over time. States favor some institutions because they are better suited to new conditions or new problems and abandonor downplay those that are not. For example, the obvious place to handle intellectual propertyrights would seem to be the World IntellectualPropertyOrganization,but the countriesthatgeneratemost patentschose to move the issue to the WTO because of it offered better enforcementmechanisms.Thus the institutionalization the issue institution was modified, but because evolved significantly, not because an older anotherone offered a better institutionaldesign.20 Even institutions that are not highly formalized and arise throughinformal and evolutionary processes may embody significant rational design principles. Sovereignty is clearly the result of historical and normativeprocesses, but at important
18. Our proposed conjectures are consistent with an evolutionary perspective that treats rational designs as superior in the sense of providing greater benefits to participants,even if participantsare unwittingbeneficiaries.Miles Kahler provides an excellent overview and discussion of the relationship institutions.Kahler 1999. The two approaches between evolutionaryand rationaltheoriesof international as and begin to align throughsuch concepts as "learning" "imitation" key factorsunderlyinginstitutional development. the 19. Of course, many efforts at institutionaldesign fail. States may misunderstand circumstances they face or wrongly anticipate how actors will respond to institutionalinnovations, or simply make mistakes. 20. See Schrader 1996.
768 International Organization
junctures (Treaty of Westphalia, Congress of Vienna, Vienna Convention) it has been the object of rationaldesign throughcodification and modification. Thus, our basic strategy is to treat institutionsas rational, negotiated responses to the problems international actors face. We can connect our definition of institutions to the language of game theory, where institutions are aspects of This equilibria,including the rules of the game and the expectationsof the actors.21 equilibriumapproachhas several importantimplications.22 First, institutionalrules must be "incentive compatible" so that actors create, change, and adhereto institutionsbecause doing so is in their interests.Consideran institutionthat can be sustained only through sanctions and whose members must apply these sanctions themselves. This is an equilibrium institution only if the members who are supposed to apply sanctions actually have incentives to do so. Incentive compatibilitydoes not mean that membersalways adhereto rules or that every state always benefits from the institutionsto which it belongs. It does mean that over the long haul states gain by participating specific institutions-or else in they will abandonthem. Second, specifying independentand dependentvariablesrequiresspecial care. An equilibrium is a statement of consistency among its elements. Decomposing an equilibriuminto causal statementsconnecting independentand dependentvariables requireslooking beyond the equilibriumitself to the sequence of, and reasons for, institutionalchanges. Third, the very institutions we seek to explain as "outcomes"may also play a causal role in shaping others, now or in the future. Consider the EU. Is it a or variable?The answerdepends on the question we "dependent" an "independent" ask and the time frame we use. If we want to explain why the EU was formed and the features it has, it is a dependent variable (by our own choice). If we want to explain the shape of some subsequentinstitution,such as the WTO or the European MonetarySystem, the EU plays a significantcausal role as an independentvariable in the institution's development. This is particularlyimportantwhen we look at which actors are relevant to a particulardesign issue. An outcome (or dependent variable)at one stage-the membershipof the EU-may become a causal factor (or independentvariable)at another-the numberof actorsrelevantin the design of the EuropeanMonetarySystem. Dependent Variables Consider an emerging internationalissue, such as global warming,the distribution of pirated software, or the sale of cloned human organs. If states want to promote a common interest,what kinds of institutionsmight they design to aid their efforts?
21. The converse is not true, and not all equilibriaare institutionsas we define them. In particular we exclude equilibriaresultingfrom tacit bargainsand implicit arrangements arise without negotiation. that 22. See Calvert 1995; Morrow 1994c; and Snidal 1997.
institutionat all. Perhaps They might first ask whetherthey need an international their national capacities are more than adequate, or they are converging on tacit arrangementsthat require little elaboration. If they could benefit from explicit cooperation,they would ask whethercurrentinstitutionscould be extended to cover the issue, in whole or in part. If the issue were novel (such as trade in cloned organs) and no existing organizationswere well suited, then diplomats, executives, scientists, policy activists, and other interestedparties might well consider creating a new organization. They would immediately confront several major questions. Should the new institution cover only cloned organs or should it also cover health- or trade-related issues? Should membershipbe limited to countries with advanced medical industries? What about other, less-developed countries?One practicalreason for being inclusive is that excluded states might evade or underminethe rules. What about including scientific institutes, biotechnology companies, health advocates, medical ethicists, and other nonstate actors? What institutionalcapacities are needed for success? Would a simple agreement suffice? Should the institutionbe centralizedto collect data, monitorcompliance, or even enforce some rules? Or should it be more decentralized,serving mainly as a forum for periodic bargaining?Should all actors be given equal voice and vote, or should some have only an informal, consultative role? What about the rules themselves in such a new and rapidlydeveloping area?Should they be clear-cutand firm, or should they be more flexible, allowing easy changes by mutual agreement or opting out by dissatisfied states? Regardless of the issue, these kinds of institutionalchoices zero in on our major institutionsdesigned as they are? To make concerns:how and why are international on these overarchingquestions,we need some clear way to markout major headway variations in institutionaldesign. The simplest solution would be to use a single or measure, one that describes institutions as, say, "stronger" "weaker."Unfortusuch measures are misleading because they collapse several important nately, institutionalfeatures into one overly simple statement. We could measure many institutionalfeaturesin great detail, yielding rich descriptionsof individualinstitutions, but this would obscurethe most important types of variationamong them. We focus on a few recurrentproblems of institutionaldesign, have chosen instead to particularlythose we can identify theoreticallyas vital aspects of cooperationand that vary in measurable ways. Our approach highlights five key dimensions:
MEMBERSHIP, SCOPE, CENTRALIZATION,CONTROL, and FLEXIBILITY. These
are not the only
importantdimensions of institutions. Others may well prove significant, theoretically and substantively.In some cases, our dimensions must be refined to clarify design issues in specific institutions. Centralization,for instance, is a broad category-perhaps too broadfor some cases. Nonetheless, our firsteffort is to reducethe myriadelements of institutionalvariationto a few measurabledimensionsthat show up repeatedlywhen institutionsare designed or modified.We now take a closer look institutions. at each dimension and consider how they vary in moder international
Membership Who belongs to the institution?Is membershipexclusive and restrictive, like the G-7's limitation to rich countries?Or is it inclusive by design, like the UN? Is it regional,like ASEAN, or is it universal?Is it restrictedto states, or can NGOs join? Membershiphas been one of the most hotly contested issues in recent years. The expansion of NATO into Eastern Europe is a key example. Expansion, for those who favor it, representsa reinvigorationof the alliance, a commitmentto the joint defense of CentralEurope,and a symbolic inclusion of new membersin the "West." For those who oppose it, NATO's movementto the East adds nothingto the defense of Western Europe and needlessly provokes an already humiliated Russia. These issues resonate widely because NATO is such a prominent and consequential institution.
Scope What issues are covered? In global trade institutions, for example, some of the toughestbattles have been over which sectors to include in negotiations.GATT left out several key economic sectors, but the WTO has expandedto incorporatemost trade issues, including agriculture and services. It may be expanded further to include cross-borderinvestments. At the other end of the spectrumare institutions like the 1965 U.S.-Canada auto trade deal designed to cover only one or two narrowlydefined issues. This agreement,too, was eventually widened when it was incorporatedinto NAFTA. Sometimes two seemingly unrelatedissues are linked. A tradeissue, for example, may be linked to a security issue to facilitate agreementand compliance. Or a side payment may be offered, as when the Nuclear Nonproliferation Treaty offered the transfer of peaceful nuclear technology to states that agreed to forgo nuclear weapons. Such side paymentsare clear evidence that scope is being manipulatedto facilitate cooperation. There is a continuum of issue coverage. At one end are institutions like the Antarctic Treaty System that cover a range of scientific, economic, and political issues. At the otherend are some early environmentalagreementsthat are restricted to a few well-defined issues, such as greenhouse gas emissions. Sometimes scope is not open to design choice because of technicalconsiderations or sharedperceptions.In the Law of the Sea negotiations,for example,jurisdiction over ocean territoriescould not be separatedfrom coastal environmentand fishing rights issues. Technological interactionsrequired that these issues be dealt with But other Law of the Sea issues seemed together in a comprehensivesettlement.23
23. A parallel and important implication within rational institutional design is that all relevant of airline "margins" choice must be considered.Barzel 1989. In John Richards'analysis of international regulationin this volume, for example, effective agreementson airline fares also requirethat airlinesbe prohibitedfrom competing on other margins, such as food quality or seat comfort.
to have little in common. Here linkage was more cognitive-a result of how issues were framed, especially under the rubricof the "commonheritage of mankind."24 One difficulty in analyzing scope is that the issues themselves are not clearly defined.Does tradein all commoditiesconstitutean issue? Or should we distinguish Although there is no general answer to this agriculturalgoods from manufactures? difficult task of assessing issue scope, focused empirical research can reveal the extent to which actors narrowor broadenthe range of mattersbeing addressed.The problem is simplified when negotiations are expanded to cover items that could clearly be dealt with separatelyor were not previously linked (as occurredwith the "baskets"of the Helsinki negotiations). Most important,changes in institutional issue linkage over time indicate changes in scope within an arrangement. Centralization Are some importantinstitutionaltasks performed by a single focal entity or not? Scholarsoften misleadingly equate centralizationwith centralizedenforcement.We use the term more broadly to cover a wide range of centralized activities. In we particular focus on centralizationto disseminateinformation,to reduce bargaining and transactioncosts, and to enhance enforcement. These categories are not exhaustive, but they cover many important centralized activities found at the internationallevel. is Centralization controversial,politically and conceptually,because it touches so directly on nationalsovereignty.According to the traditionalview, states reject any form of centralized internationalauthority. Internationalrelations is seen as an immutableanarchy.This is a powerful assertion,but it is only partlyright. It blends a simplifying assumption(thattheory building should begin with states as independent units) with some hyperbole and errantconclusions. States understandablyguard their domestic authority and their control over foreign policy. They are suspicious of encroachmentsby other states and strongly bodies. But saying that resist any shift of sovereign responsibilitiesto superordinate states rarely devolve such authorityis inaccurate,and it is a misleading basis for constructingtheory. After all, Europeanstates not only signed the Treaty of Rome but also agreed to the Single EuropeanAct, which permitsmajorityvoting.25They went still furtherat Maastricht,when they abolishednationalcontrolsover money.26 institution,but centralizedcontrols The EU is uniquely powerful as an international are importantelsewhere. The dispute-resolutionpanels of the WTO are a particularly significant example. The least intrusive form of centralizationis informationcollection, and many internationalinstitutionsengage in it. Members of the IMF, for instance, need not
24. Haas 1980. 25. Moravcsik 1991. 26. See Kenen 1995; and Moravcsik 1998.
gather their own data on others' balance of payments. Instead the IMF regularly collects, evaluates, and publishes itemized statistics on its members' payments. Bargainingproceduresand rule enforcementcan also be more or less centralized. At the WorldBank, for instance,specialistsnegotiateloans for economic adjustment or major infrastructureinvestments. These packages require collective approval from a centralized body of members. Most internationalorganizationshave relatively decentralizedenforcementarrangements. They specify possible punishments for rule violations but leave it up to the members to apply them. Because these multilateralsanctionsare both limited and well specified,they minimize the chances for disproportionate punishmentor cycles of retaliation.Still, the members themselves must apply the decentralizedpunishmentsand bear the inevitable costs. GATT (and now the WTO) have relied on such decentralized sanctions for decades. If a disputepanel found violations of international traderules, it was up to the injuredpartyto retaliatewithin specified limits. GATT itself had no centralized power to punishor reward,only to authorizeindividualmembersto do so. This also shows how internationalorganizationscan combine elements of centralizationand decentralization. The WTO's centralizedarrangements judging tradedisputesgo for hand-in-handwith decentralizedarrangements enforcing the judgments. for
Control How will collective decisions be made? Controlis determinedby a range of factors, including the rules for electing key officials and the way an institutionis financed. We focus on voting arrangements one important observableaspect of control. as and Even if membershipis universal,some statesmay carryconsiderablymore weight than othersbecause of voting and decision-makingrules. Two interrelated rules are whetherall membershave equal votes and whethera minority especially important: holds veto power. If a minority can veto, its votes inherentlycarry special weight. In the UN GeneralAssembly all membershave equal votes. In the SecurityCouncil they do not, since only the permanentmemberscan veto resolutions.The IMF and World Bank have explicit weighted-voting rules; the larger economies, which votes. Another element provide capital to these institutions,carry disproportionate of control is whethera simple majority,a super-majority, unanimityis required. or If a super-majority needed, some state (or combinationof states) may be able to is block new rules, members, or officers. Finally, we distinguish control from centralization.While centralization may reduce control in some cases, the two dependentvariablesgenerally vary independently. For example, changes in the voting rules within a quasi-legislative component of an international institutionrepresentchanges in controlthat do not affect the level of centralization.Similarly, centralizing information collection usually has little, if any, effect on who controls an institution.
Flexibility How will institutional rules and procedures accommodate new circumstances? Institutions may confront unanticipated circumstances or shocks, or face new demandsfrom domestic coalitions or clusters of states wanting to change important rules or procedures.What kind of flexibility does an institutionallow to meet such challenges? It is important to distinguish between two kinds of institutional flexibility: adaptive and transformative."Escape clauses" are a good example of adaptive flexibility. They allow members to respond to unanticipatedshocks or special domestic circumstanceswhile preserving existing institutionalarrangements.The general goal is to isolate a special problem-such as a spike in steel imports from a few producing countries-and insulate the broaderinstitution (in this case, the GATT/WTO) from its impact. This limited flexibility is designed to deal chiefly with outlying cases, to wall them off from run-of-the-millissues. to themselves in ways that Some institutionshave built-inarrangements transform are more profound. This deeper kind of flexibility usually involves clauses that permit renegotiationor sunset provisions that require new negotiations and ratification for the institutionto survive. The initial terms of commodity agreements,for example, are typically five to seven years, after which they expire and have to be renegotiated.GATT did not have such a provision, but its periodic roundsof trade negotiations facilitated planning for larger institutional changes, leading to the WTO. GATT's existing rules did nothing to block these larger changes, and its regularforums served to promote them.
Independent Variables
To explain variationin institutionaldesign, we focus on the following independent enforcement problems (ENFORCEvariables: distribution problems (DISTRIBUTION);
and uncerMENT); number of actors and the asymmetries among them (NUMBER);
tainty about behavior, the state of the world, and others' preferences(UNCERTAINTY
ABOUT BEHAVIOR, UNCERTAINTYABOUT THE STATE OF THE WORLD, and UNCERTAINTYABOUT PREFERENCES).
Enforcementof agreementsis a cornerstoneconcernin international anarchy.But recent debates have increasingly stressed that to understandwhich, if any, internaissues. The tional institutionalbargainsare struck,one must examine distributional numberand relative size of key actors has been a long-standingconcern in debates aboutinternational cooperation,hegemony, and, more recently,the interrelationship of regional and global politics. Finally, uncertaintyis the linchpin of traditional security problems and is equally central in economic and environmentalissues. These variables also play a crucial role in game theory. Enforcement and distribution problems emerge in any strategic situation. Number is the central variableof collective-action theory, and we broadenit here to include explicitly the
774
asymmetriesthat are so importantin internationalaffairs. Finally, many important theoretical developments in game theory over the past two decades center on uncertainty. Since we extend the existing tradition of cooperation theory, it is useful to compare our independentvariables with Oye's.27 After all, institutionsto promote cooperation must be designed around the factors that affect cooperation. But we adapt the independent variables to address the particular questions raised by institutionaldesign. Oye focuses on three independentvariables. The most important is "shadow of the future."We do not focus on this as a primary source of institutionalvariationbecause the general conditions for cooperationare typically met undercontemporary conditions of high interdependence.28 Instead,we emphasize how variation in the significance of enforcement problems across different issues affects institutionaldesign. Oye's second independent variable is the type of 2 X 2 game being played, though with an emphasis on Prisoners' Dilemma. Simple games have yielded importantinsights and have been subjected to importantcriticisms.29The most on importantsubstantivecriticismis thatconcentration Prisoners'Dilemma leads to an overemphasison enforcementand cheating and to an underemphasison distributional conflicts.30This problem can be partially solved by shifting attentionto another2 X 2 game (Coordination,for example), but each new game misses some other salient problem (such as enforcement). We resolve this by looking at distributionproblems as a second independentvariable.31 We use a broaderversion of Oye's thirdvariable,"number." Looking beyond the raw numberof actors relevantto an issue, we include asymmetriesthat might exist among them due to different capabilities. This considerationwas importantin the hegemony literatureand becomes even more so in understandinghow differentsized actors share control in institutionalizedcooperation. Finally, and most important,drivenby advances in the economics of uncertainty and game theory we add "uncertainty" a new category of independentvariable. as can impede cooperation, but its impact can be managed through Uncertainty institutions. Indeed, one feature common to our independent variables is that
27. Oye 1986. 28. Alternatively,states will not waste time designing institutionsthat will not be enforced by their own incentives. 29. In particular,once the games are complicated even slightly, the clean distinctions among them break down. When Prisoners' Dilemma repeats throughtime, for example, multiple equilibriaemerge, and the supergamecontains distributionalproblems. Similarly, recurringBattle of the Sexes problems create incentives for some states to shift the prevailing equilibrium. 30. See Krasner1991; and Grieco 1988. 31. James Fearon makes a parallel argumentthat, at a sufficiently general level, all problems in internationalrelations have a common strategic structure.Fearon 1998. States must choose among the range of available cooperative arrangementsand ensure that participantswill adhere to the chosen We label these the "distribution arrangement. problem"and the "enforcementproblem,"respectively.
game-theoretic logic allows us to connect them to the dependent variables of institutionaldesign.32
DistributionProblems When more than one cooperative agreementis possible, actors may face a distribution problem. Its magnitudedepends on how each actor compares its preferred alternative to other actors' preferred alternatives. In a pure Coordinationgame, where both actors prefer the same coordinationpoint(s), there is no distribution problem. Distribution problems are greater when actors want to coordinate in a "Battle of the Sexes" game according to the intensity with which they prefer alternative coordination points. In Prisoners' Dilemma games where there are multiple efficient equilibria,the distributionproblemdependson actors' differences Finally, the problem is most severe in a zero-sum "along the Pareto frontier."33 because a better outcome for one leaves less for the others. game Distributionproblems are closely relatedto bargainingcosts.34 In general, where the distributional implicationsof a choice are small (such as when only one efficient outcome is possible or the shadow of the future is short), bargainingcosts will be implicationsare large (such as relatively small. In situationswhere the distributional when there are multiple, substantiallydifferentefficient outcomes or the shadow of the future is long), bargainingcosts will likely be large. Distribution problems interact with the other independent variables, but they should be kept separate.Most important,distributionproblems are not the same as Uncertaintyarises when an actorcannotanticipatethe outcome that will uncertainty. result from an agreementand knows only the stochastic "distribution" generating the outcome. In their collaborative venture to develop an anti-missile system, for example, Japan and the United States are uncertainwhether the research will be successful even though they are sure they will both share fully in the findings. In contrast, a distributionproblem refers to selecting one outcome from a range of known possible outcomes. In allocating quotas for harvestingWest Coast salmon, for example, Canadaand the United States know the total numberof fish that will be caught; the problem is determiningeach country's allotment. Of course, these problems intertwine in many situations where actors choose among agreements characterized differentstochasticdistributions.This is trueof fishing agreements by over time where both the allotmentsbetween states and the size of the fish harvest over time are at stake.
32. We asked contributorsto examine these independentvariablesbut also invited them to consider others; thus the project as a whole is open to a wider set of independentvariables, albeit in a more inductive way. 33. Krasner1991. 34. Fearon 1998.
776
EnforcementProblems Enforcement problems refersto the strengthof individualactors' incentives to cheat makes everyone better on a given agreementor set of rules. Even if an arrangement off, some or all actors may prefer not to adhere to it because they can do better individually by cheating-the heart of Prisoners' Dilemma and public goods problems. The enforcementproblem arises when actors find (current)unilateralnoncooperationso enticing that they sacrificelong-termcooperation.It can be measuredby the minimum discount factor (a state's valuation of future, as opposed to current, benefits) necessary to supportcooperation. Seen this way, the necessary discount factor is a characteristicof the issue-including actors' payoffs from cooperation and defection and how frequently they interact-but not of how much actors actually value the future. Issues where actors have large incentives to break an agreement require higher discount factors to support cooperation than do issues where the immediate gains from noncooperationare smaller. Although we focus on settings of high interdependencewhere cooperation is generally possible, there is significant variationacross issues. At one extreme are cases with no enforcementproblems, such as agreementsto set technical standards where actors have no incentive to defect. Within the context of repeatedPrisoners' Dilemma games, self-enforcing agreements may arise if incentives to defect are small relative to the shadow of the future.But if incentives to defect are greater,or interactionsare less frequent,enforcementproblems emerge. Most situationscontainboth distributionand enforcementproblems.In efforts to halt stratospheric ozone loss, for example, the ozone regime needed to set targetsfor chloro-fluorocarbon (CFC) emissions and establishrules for cutting reducingglobal back CFC productionand use. Differentrules obviously impose quite differentcosts on variousstates.Whateverrules are chosen still have to be enforced.Knowing this, states may choose particularrules partly because they are easy to monitor and enforce. In this way problemsof distributionand enforcementare tightly connected. Distribution and enforcement can be blended in differing proportions. Some problems are more squarely related to enforcement, with distributionalconsiderations clearly secondary.If first strikescan paralyzeone's opponent,enforcementof any arms control agreement overwhelms any distributionalconcerns about armament levels. Otherissues presentmajordistributionproblems,with enforcementas a secondaryissue. Macroeconomiccoordinationamong the G-7 countriesseems to The same could be said of the last three GATT rounds. The have this property.35 critical issue was who would make what concessions, not whether the resulting agreementswould be enforced. Separating enforcement problems from distribution problems is an analytic choice, not a substantiveclaim. Unlike early work based on Prisoners'Dilemma or more recent work based on Coordination,it enables us to consider the more typical
35. Webb 1991.
case, where enforcement and distributionproblems occur simultaneously.It does not capturemore nuanced interactionsbetween enforcementand distributionproblems, but by first examining the institutionalissues raised by these "maineffects," we will be better situated to understandthe others. Finally, it is necessary to keep enforcement problems distinct from the other independentvariables. Uncertainty and large numbers usually aggravate enforcement problems, but enforcement problems can arise even in repeated-gamesituations with small numbers and no uncertainty. Number of Actors Number of actors refers to the actors that are potentially relevant to joint welfare because their actions affect others or others' actions affect them. Sulfur emissions from factories in the U.S. Midwest, for example, cause acid rain in EasternCanada and New England,an issue involving two countries.Greenhousegases emittedfrom the same factories contribute to global warming, an issue affecting more actors because of the large-scale consequences of global climate change. If firms are seen as the relevantactors,then the numberof actorsis significantlylargerin both cases. The numberof actors involved in militaryissues depends on technology and on states' ability to harmor help one anothermilitarily.Peace in the Middle East now depends on more states than it once did because technological innovations have increasedthe range of militaryaircraftand thus the numberof states that can affect the military balance. Were Pakistanable to target Israel with nuclear weapons, it, too, would become a key actor. Number does not depend solely on geographic or technological factors and is For example, a often determinedby prior political and institutionalarrangements. decision by the EU about monetary union is effectively a fifteen-state decision, regardlessof its effects on outsiders,because EU membersmade a political decision to limit the numberof states involved in the process, not because other states are unaffected.Similarly,when NAFTA takes up an issue, only its three membershave a voice, whereas the same issue taken up within an expanded hemispheric trade would involve more states. In effect, the priorinstitutionalmembership arrangement decision has redefined the range of "potentiallyrelevant" actors for the issue at hand.36 to Thus it is important distinguishbetween the independentvariable,number,and the dependentvariable, membership.Number is an exogenous feature of the issue context, including prior institutionaldevelopments, in which an institutionmay or may not be established. It includes the set of interested actors and their relative power in and importanceto the issue. In contrast, membershipis an endogenous design choice made in the course of establishing, changing, and/or operating the institution.It includes, for our purposes, the rules governing who is a member and
36. Snidal 1994.
(if relevant)differentclasses of membership.Over time, priormembershipchoices may affect number-that is, endogenous choices become exogenous constraintsbecause institutionalsettings, such as the EU or NAFTA, determinewhich actors will have standingin subsequentinstitutionalnegotiations. Number also includes asymmetricaldistributionof actors' capabilities.On some issues many states may be nominallyinvolved, but only a few really drive the issue. Every state has an interestin the international economy, for example, but few have the economic power to determineits course. Similarly, many states produce some oil, copper,or bauxite,but only a few states dominatethe global productionof each. The actors involved in an issue are not always the same as those who become members of the final institution. Although the entire EU membership discussed and monetaryunion, only some met the requirements chose to join. Similarly,while trade affects virtually all states, not all have played an active role in multilateral negotiations, and not all are members of the WTO. Uncertainty Uncertaintyrefers to the extent to which actors are not fully informedaboutothers' behavior, the state of the world, and/or others' preferences. These distinctions correspondto three importantelements of any strategic situation:choices, consequences, and preferences,respectively;and they may have differentimplicationsfor institutional design. For example, uncertaintyabout behavior makes cooperation more difficultin many cases, but uncertaintyaboutthe state of the world may, under certainconditions, make cooperationeasier. Therefore,our assertionsare not about generic effects of uncertaintybut about the differentways states design institutions to cope with specific types of uncertainty. Uncertainty about behavior. States may be unsure about the actions taken by others.If states agree not to pursuetechnologies associatedwith the developmentof biological or chemical weapons, for example, some states may have no way of knowing whetherothers are abidingby the agreement.Similarly, if countriesagree to restrict sulfur emissions to reduce acid rain, how can they be sure others are complying with the agreement?37 Uncertainty about the state of the world. Uncertaintyabout the state of the world refers to states' knowledge about the consequences of their own actions, the actions of other states, or the actions of internationalinstitutions. This could be scientific and technical knowledge or political and economic knowledge. Consider the dispute over the SpratlyIslands, which lie off the southerncoast of China and have been claimed by a number of states. Any agreement governing the dispute would have to take into account that no one knows how much oil is actually there or its future value.
37. Levy 1993.
Governments are often unsure what their Uncertainty about preferences. counterparts really want. We assume states know theirown preferences,but they are often uncertain about the preferences or motivations of others. A key problem underlyingarmscompetitionis determiningwhetheranotherstate is simply seeking its own security or is greedy and expansive. Does India's nuclear testing reflect a desire to aggrandizeitself at Pakistan'sexpense or to defend itself againstChina?Of course, a majorproblem in determiningothers' preferencesis that states may have incentives to misrepresenttheirpreferences,either verballyor throughtheir actions. We do not use standardgame-theoreticterminology, such as imperfectinformation or incomplete information,because it would obscure importantdistinctions.38 For example, we could capture uncertaintyboth about the state of the world and about preferences(or type) throughgames of incomplete information.But collapsing these into one category prevents us from drawing nuanced inferences about institutional design. Foreshadowing the conjectures discussed later, membership aboutpreferencesbut not aboutthe stateof the world. rules may mitigateuncertainty flexibility provisions can help states cope with uncertaintyaboutthe state Similarly, of the world but have no effect on reducing uncertaintyabout behavior. Although distinguishingamong these kinds of uncertaintyis useful conceptually, in practice they are often combined. For example, do Europeanefforts to restrict imports of U.S. beef produced with hormone supplements reflect a concern for consumers' health or for local farmers' profits (uncertaintyabout others' preferences)? Scientific uncertainty(uncertaintyabout the state of the world) was also panel ruledthathormones presentinitially but was resolved when a WTO-appointed no health threat.An obvious solution would be to label importedbeef as such posed concernsabout and let individualEuropeansmake theirown choices. Unfortunately, such a labeling system (uncertaintyaboutbehavior)would frustratethis monitoring solution. Different mixes of uncertainty often characterize an issue. For example, the environmentalarea is plagued by enormousuncertainty(most of it scientific) about the state of the world and much less uncertaintyaboutpreferences.In contrast,there was little uncertaintyabout force structuresduringthe latteryears of the Cold War, but each superpowerhad significantuncertaintyabout the preferencesof the other. We would expect the design of agreementsin these areas to reflect their different circumstances. InteractionsAmong IndependentVariables Ourresearchdesign is quite simple. We have isolated a set of independentvariables institutionaldesign featuresthatwe expect will determinethe choice of particular In our conjectures,we focus on "maineffects"-that is, the our dependentvariables. bivariaterelationshipsbetween the independentand dependentvariables.
38. We do adopt standard terminologyin using the term uncertaintyinstead of risk. See, for example, Kreps 1990; Hirshleiferand Riley 1992; and Osborne and Rubinstein 1994.
780
This approachhas several advantages.It providesa generalframeworkfor a wide range of empirical studies and fosters comparisons across cases while allowing individual analysts to explore the implications of interactions in their particular cases. Moreover, the emphasis on bivariaterelationshipsallows us to connect our conjecturesclosely to existing theoreticalwork-which would be possible for some but not all of the more complex interactions.Although simplicity has tremendous advantages, it ignores potential interactions among the independent variables. Enforcement problems may be combined with uncertaintyabout preferences or actions, as in an arms control context. Or distributionproblems may be combined with large numbers, as in environmental public goods contexts. Because our independent variables may combine in many ways, we need to consider the For example, when an enforcement problem significance of their interactions.39 occurs in a repeatedPrisoners'Dilemma, cooperationis possible providedactorsare sufficiently patient. But when uncertainty about actions enters the picture, the viability of cooperative strategies declines, since these strategies hinge on actors' knowledge of each other's behavior. Here the combination of two problems is substantiallyworse than either one alone. Similarly, uncertaintyabout the state of the world can interactwith distributional problems,making cooperationeven more challenging.4 The interactionof independentvariables can also enhance cooperation. While both large numbers and distributionaldifferences typically impede cooperation, sometimes large numbersmitigate distributional problemsby easing relative gains concerns or by offering additionalways to balance costs and benefits across actors.
Conjectures About Rational Design

In this section we develop a series of conjecturesthat connect our independentand dependent variables. We call these "conjectures"to indicate that they represent generalizationsbased on a common rational-choicetheoreticalframework,although they are not formally derived here; however, in presentingthe underlyinglogic of each conjecture we identify close variants that have been formally derived by scholars working in the rational-choicetradition.Although the conjecturesfollow from this general framework, individual conjectures depend on logics that may entail specific substantive assumptions. For example, public goods arguments assume that all actorssharethe same goals, whereas"screening" argumentssuppose
39. Interaction effects may be positive, negative, or zero-that is, when two "problems" arise together in a given context, theirjoint effect may be less thaneitherproblemindividually(a large negative effect) or more than either problem individually but less than the sum of the two (a small negative effect). Alternatively,the combined effect may equal the sum of the two individualeffects (a zero interaction effect) or be greaterthan the sum of the individualeffects (a positive interactioneffect). 40. Koremenos 1999a.
that some actors do not.41Thus the conjecturesneed not be fully consistentwith one another in this sense. Similarly, not all conjectures will apply to every casesomething we leave to the individual case studies to determine. In the volume's conclusion we discuss the empirical and logical relationships among the conjectures. We now address four broad assumptionsthat underlie our conjectures. 1. Rational design: States and other internationalactors, acting for self-interested reasons, design institutionspurposefullyto advance theirjoint interests. We thus make standard assumptions:actorshave (well-behaved)preferencesover various goals; and the pursuit of those goals is guided by their beliefs about each others' preferences and the relative costs and benefits of different outcomes; and actors are constrainedby their capabilities.42Although the process of institutional design is usually contentious, we do not focus on the bargaining among the of but participants on the broadcharacteristics the institutionaloutcomes they select. do not simply reflect the preferencesof the individual actors but These outcomes rather represent their joint efforts-and "compromises"among their preferences-to improve their equilibriumoutcome given the strategic circumstancesthey face. That is to say, they concern the equilibrium outcomes that result from the strategicinteractionof states, each of which has preferences.Of course, for certain sets of preferences (such as when distributionalissues are absent), the strategic aspects of states' interactionare trivial, and institutionaldesign outcomes appearto reflect only preferences. 2. Shadow of the future: The value of future gains is strong enough to support a cooperative arrangement. Actors have a sufficiently high density of interaction-and a sufficiently high discount factor-that cooperationis potentiallysustainable.We take a long shadow of the future to be a general condition of contemporaryinternationalinterdependence, but one subject to considerable variation across issues. On some issues, actors may not interactwith sufficient frequency for future incentives to be strong enough to support cooperation by themselves.43On other issues, such as peacekeeping, unilateral incentives to defect or distributivedifferences may make cooperation difficult. A variety of other circumstances-especially uncertaintyand large numbers-may make cooperationnot only difficultto achieve but also difficult to enforce. Therefore, general internationalcircumstances may be propitious for cooperation,but the particularcircumstancesin any issue may be problematic.
41. We thank Jim Morrow for this example, which correspondsto a comparisonof conjecturesM1 and M2. 42. We focus on states as key actors,thoughmost of the analysis can be generalizedto nonstateactors. 43. Of course, harsherpunishment strategies can be used to support greater cooperation when the shadow of the futureis short;however, such strategiesare subjectto problemsof renegotiationproofness. See Downs and Rocke 1995; and Abreu, Pearce, and Stacchetti 1986.
782
3. Transactioncosts: Establishingand participatingin internationalinstitutions

is costly.44
When creatinginstitutions,states need, for example, to acquireinformationabout the issue, about each other, and about the likely effects of alternativeinstitutional forms. One way they do this is through negotiations. There are other types of transactioncosts as well, such as safeguards to ensure compliance and sustain cooperation.45As David Lake explains, these safeguards may include sanctions, hostages, and dispute-resolutionarrangements.46 An importantaspect of our independentvariablesis that they may raise or lower transactioncosts. For example, the largerthe numberof actors,the slower and more cumbersome the negotiations. Likewise, greater uncertainty may make it more costly to write complete contractsto deal with every contingency.Thus, numberand uncertaintyoperate partly throughtheir impact on transactioncosts, which is why we separateout such costs in our assumptions.We focus on these variablesrather than on transactioncosts directly because they are more readily observable. 4. Risk aversion: States are risk-averseand worry about possible adverse effects when creating or modifying internationalinstitutions. Risk-averse actors prefer a certain outcome to a chancy one when each has the same expected value. This assumption is the bedrock of moder realism, where states' fears of destructionand keen interest in preservingtheir sovereignty dominate their strategic calculations. However, even realist states may trade off some sovereignty if they reap large enough gains in return.47Institutionalistshave a broader view of what states value, but they, too, typically assume states are risk-averse. With these four assumptionsin mind, we now turnto specific conjecturesabout international institutional design. Because our primary purpose is to generate testable propositionsthat will guide the empirical analysis of internationalinstitutions, we frame the conjecturesin a general way. Each conjecture addresses the expected effect of a change in a particular independent variable, such as the level of uncertainty or the severity of the distributionproblem, on one of our dependentvariables. Thus our logic is that of comparative statics-that is, we ask how a (perhaps hypothetical) change in an independentvariablewill affect the equilibriuminstitutionaldesign. For example, if uncertaintyabout the state of the world increases, will states design more or less
44. For a general discussion of transactioncosts, see Williamson 1985. For an importantapplication to internationalpolitics, see Lake 1996. Unlike Williamson, we do not assume that the presence of transaction costs implies bounded rationality. Transaction costs refers to the costs of making an agreementand operatingit, not of doing what the agreementis designed to do (for example, if two states agree to jointly build a dam, the costs of negotiating and administeringthe agreementare transactions costs, but the costs of building the dam are not). 45. See Williamson 1985; and Yarbroughand Yarbrough1992. 46. Lake 1996. 47. Morrow 1991.
Institutions 783 The Rational Design of International
flexibility into an internationalinstitution?In answering this question, we assume that everythingelse remainsconstant.We emphasize the "maineffects" of individual independentvariables ratherthan more complicated interactionsamong them. These simplifying assumptions are necessary given the level of theoretical and empirical generality to which we aspire. After presenting the conjectures we will discuss the limitations of both comparativestatics and main effects approachesin terms of design interactions. ConjecturesAbout Membership Membershiprules determine who benefits from an institution and who pays the costs. They work in several ways beyond simply reducing or enlarging size. By setting criteriafor inclusion, for example, they affect the group's homogeneity and asymmetries.Not surprisingly,such rules have importantconsequences for interactions. Conjecture Ml:
WITH THE SEVERITYOF THE INCREASES MEMBERSHIP RESTRICTIVE
PROBLEM. ENFORCEMENT
The more severe the enforcementproblem, the more restrictedthe membership. When actorsface an enforcementproblem (thatis, when individualsdo not have an incentive to voluntarilycontributeto group goals), collective action is problematic. Moreover, the severity of the enforcement problem increases with the number of For actors, as MancurOlson demonstrated.48 this reason, Oye argues that reducing multilateralinteractions to bilateral ones will increase the incidence of cooperation.49 The literatureon "clubgoods" shows that a less drasticreductionin membership may be effective in promotingcooperationamong somewhat larger groups.50If an institutional arrangementrestricts the benefits of cooperation to members, actors have an incentive to pay the price of admission to the club. One of the most importantfeatures of institutions is to define these boundaries of membership.51 Furthermore,when uncertainty about a state's capacity to comply is at issue, inclusive membership may be suboptimalbecause, as George Downs and David Rocke argue, "every time the third state violates the treaty,the other two states are forced to suspend the cooperationbetween them to punish it."52
48. Olson 1965. 49. Oye 1986. Pahre points out that under strict public good conditions, such restrictions are the suboptimal.Pahre 1994. He demonstrates possibility of large-nmultilateralcooperationundercertain conditions.But unlike conjectureM1, his equilibriumis vulnerableto bad information,and it needs other institutionalsupportsthat we discuss under conjecturesC1-C3. 50. Buchanan 1965. 51. Snidal 1979. 52. Downs and Rocke 1995, 126.
The effectiveness of membershiprestrictionsdepends on the specific characteristics of the issue. In issues like CFC emissions, for example, preventingfree riding is virtually impossible. Alliance guarantees, however, are usually effective in restrictingnonmembersfrom receiving securitybenefits. Enforcementis not always a problem, of course. Agreements on internationalstandardsare a good example. Under preference configurationslike these, where everyone benefits from wider participation,free riding and enforcementare not issues, and membershiptends to be inclusive.
Conjecture M2:
PREFERENCES. RESTRICTIVE MEMBERSHIP INCREASES WITH UNCERTAINTY ABOUT
Membershipenables states to learn about each others' preferencesif the membershipmechanismcan distinguishcooperatorsfrom noncooperators. Ideally, a state that values the goals of an organizationwill want to join, whereas one that wants a free ride will find it too costly to join a regime they intend to violate. In formal terms,membershipis a costly signal. Effective membershiprules createa separating equilibriumwhere only those who share certain characteristicswill bear the costs necessary to be included in an equilibrium.53 The WTO, for example, requires prospective members to bring key domestic economic rules in line with WTO rules-perhaps with phase-in allowances or special considerationsfor certain categories of states. Similarly, NATO will not accept a new member until it meets certain domestic political requirementsand brings its militaryup to certainagreed-uponlevels. By requiringconcessions, these organizationsensure that prospective members are willing to bear the necessary costs and are likely to be cooperatingmembersdown the road.When the adjustment of membershipis too low, membershipis not informative. price When membershiprules are a significanthurdle, they say something significant about nonmembersas well. Refusal to sign the Nuclear Nonproliferation Treaty is a strong and clear signal to other states. Again, it is interestingthat states unwilling to commit to this regime generally choose not to sign the treatyratherthan to sign but disobey.
Conjecture M3:
INCLUSIVE MEMBERSHIP INCREASES WITH THE SEVERITY OF THE
DISTRIBUTION PROBLEM.
Realists argue that states care not only about their direct outcomes from cooperative interactionsbut also how well they fare comparedwith others.54These distributionalor relative gains concerns create zero-sum considerationsthat seri53. Spence 1974 illustrates how education provides a costly signal of the quality of prospective employees to employers. Spence 1974. Fearon applies signaling models to crisis bargaining.Fearon 1994. See also Kydd 2000a,b. 54. See Waltz 1979; and Grieco 1988.
the ously impaircooperationin bilateralsituations.One remedyis to rearrange terms of cooperationso that benefits are more equally balanced,but this may be difficult or costly. An alternativecapturedin this conjectureis to expandthe numberof states involved in the issue because the zero-sum properties are rapidly attenuatedas membershipincreases.55 Including additional members may also mediate distributional problems by expanding the possibilities for tradeoffs among the members. Thus an agreement might give state X the short end of the stick comparedwith state Y but compensate stateX with the long end of the stick comparedwith state Z and so forth.This is one advantageof multilateraltrade agreements.Such possibilities often occur because new membersimplicitly increasethe range of issues included (for example, tradable products).We deal with these considerationsin the next section on issue scope. ConjecturesAbout Scope units. Instead,they are constructed International issues do not come as pre-packaged and evolve in complicatedways. While the resultingissue scope partlyderives from technological, cognitive-ideational, and other factors that are not analyzed here, rationalinstitutionalanalysis can explain key patternsof linkage within institutions. We focus on the deliberatechoices states make about which issues to include in an institutional framework. In particular,when do states bring together issues they might otherwise have dealt with separately? Our first conjecture follows from efficiency considerations: Conjecture S :
AMONG ISSUE SCOPE INCREASESWITH GREATERHETEROGENEITY
LARGERNUMBERSOF ACTORS.
When states are similarly positioned on an issue, they share common interests over a collective international policy (if any is needed), althoughthey may well have difficulties achieving that policy. Moreover, their relative symmetry on the issue may suggest a focal resolution,especially that all adopt a similarnationalpolicy. In these cases an issue often resolves on its own. As the numberof actors increases, however, the heterogeneitywithin the group will typically also increase. This is especially likely in internationalsettings where the additional actors are often qualitatively different from earlier actors (for example, less-developed countriesjoining a group of developed countries).5
55. Snidal 1991. 56. We do not claim thatheterogeneitypromotescooperation;in some cases it promotesdistributional differences and conflict. Our position is that linkage provides an institutionalmeans to harness these differences in a mutually beneficial way. Also, having a larger number may promote heterogeneity in capabilities(which we do not addresshere). For an insightful discussion of these points that also relates heterogeneityto institutionaldesign, see Martin 1994.
786
When actors have heterogeneous interests, issue linkage may generate new for opportunities resolving conflicts and reachingmutuallybeneficial arrangements. James K. Sebenius demonstrateshow adding issues "can yield joint gains that enhance or create a zone of possible agreement."57 The paradigmaticexample is "gains from trade,"both in the limited sense of exchanging commodities and in the broadersense of connecting issues. When one actor values issue X more than issue Y, and the other ranks them the opposite way, both can be made better off by exchange, that is, by agreeing to defer to each other on these issues. Environmental issues that are importantto postindustrialstates, for example, are often linked to issues of developmentand technology when less-developed states with less intrinsic interest in environmentalquality are essential to the arrangement.58
Conjecture S2:
PROBLEM. ISSUE SCOPE INCREASES WITH THE SEVERITY OF THE DISTRIBUTION
Linkage not only allows states to increase efficiency but may also allow them to overcome distributional obstacles.59When the benefits of an issue accrue primarily to a few, and the costs fall disproportionately others,linkage to anotherissue with on different distributionalconsequences allows cost-bearingstates to be compensated by those who reap the gains.60When each state cares relatively more about one of two issues, linking the negotiations may be the mutually preferredoption.61 In particular,the more each state cares about "its" issue, the more essential linkage becomes in an agreement.HowardRaiffa makes an even strongerassertion,arguing thatincreasedscope can transforma zero-sumgame with no zone of agreementinto a positive-sum game.62 Conjecture S3:
PROBLEM. WITHTHE SEVERITY THE ENFORCEMENT OF ISSUE SCOPEINCREASES
57. Sebenius 1983, 314. 58. In some cases, membership may act as a mediating variable through which number affects endogenous variables such as scope. Even in such cases, numbermay also have direct effects, perhaps due to asymmetriesamong the parties, for which member is not a mediating variable.This complexity is typical in a system with multipledependent(or endogenous)and independent(or exogenous) variables. Our conjecturesfocus on the impact of individualindependentvariables' main effects and thus hold the other independentvariables constant, but not the other dependentvariables. 59. Tollison and Willett 1979. 60. ConjecturesS1 and S2, though distinct, share a similar logic. In each case differences among the actors lead them to expand the issue set in order to find a better outcome. In this way, distributional differences (which cause conflict within issues) are the engine of efficiency gains (across issues). For an instructiveanalogy in the social-choice literatureon logrolling, see Mueller 1989. Logrolling, however, occurs within an institutionalframeworkand thus can lead to Pareto-inefficient moves. Riker and Brams 1973. We would not expect this in the design of new institutionalarrangements. 61. Busch and Koremenos 2001a. 62. Raiffa 1982.
When the incentives on an issue are insufficient for decentralizedenforcement, The logic here is the same as in linkage to other issues can provide enforcement.63 the shadow of the futureconjecture,except that this works across issues ratherthan over time. The United States might be unableto resist domestic pressuresto impose tariffson Europeanwine, for example, were it not for the realizationthat such action would invite retaliationfrom the Europeanson U.S. beef. Lutz-AlexanderBusch and BarbaraKoremenosshow formally that the higher the discount rate requiredto supportcooperation(thatis, as the enforcementproblemis more severe), the greater the probabilityof issue linkage.64 Since all three conjectures point to advantages of greater scope, the question naturallyarises, Why isn't everythinglinked to everythingelse? The answer is that increased scope also has costs. These include the extra bargainingcosts associated with additionalissues and the greaterprobabilitythat some actor will "holdup" the agreementto gain additionalbenefits.65The risk of unraveling,whereby failure in one issue may lead to failure in all linked issues, is also greater. What our conjecturespredictis that, all else equal, as the independentvariablesincrease, the marginalbenefits of additionalscope exceed the marginalcosts. This leads rational states to increase scope until the marginal cost of adding another issue roughly equals the marginalbenefit. ConjecturesAbout Centralization Internationalinstitutions can be centralized in a variety of ways. An international capacities,for example, without agency may have centralizedinformation-gathering having centralized adjudicativeor enforcement capacities. In the conjectures that follow we emphasize general tendencies of centralization rather than specific combinations.
Conjecture C1:
CENTRALIZATION INCREASES WITH UNCERTAINTY ABOUT BEHAVIOR.
The Folk theoremholds that when states interactover extended periods they can achieve cooperative outcomes on a decentralizedbasis through strategies of reciprocity. But when states are uncertainabout others' behavior, they cannot achieve the same mutuallybeneficialoutcomes. Greaternoise lowers the joint gains they can achieve.66Downs and Rocke show how tacit bargainingand trigger strategiescan make the best of this situation.67 However, centralizedinformationmay offer a more
63. See Hardin 1982; McGinnis 1986; and Bemheim and Whinston 1990. A more nuancedversion of this conjecture would consider the interrelationships among the issues, for example, whether they are substitutesor complements. See Spagnolo 1997. 64. Busch and Koremenos 2001a. 65. Thus our independentvariables may affect the costs as well as the benefits of scope. 66. Kreps 1990. 67. Downs and Rocke 1990.
effective alternativeif it can reduce uncertaintyaboutbehaviorto make (otherwise) decentralizedcooperationmore effective.68 The law merchant model illustrates the value of centralizationin promoting The law cooperationwhen agents are uncertainaboutone another'spast behavior.69 merchant system includes a centralized actor who serves as a repository of informationaboutthe past performance traders.This actormakes the information of availableto prospectivepartners, bond thatfacilitates therebycreatinga reputational current transactions. This actor plays a further centralized role in adjudicating disputes and awardingdamages as warranted. Centralizedinformationnot only lets states know how others have behaved but of also can provide valuable interpretations that behavior. States will know better whether others' noncooperationis intentionaland deserves retaliationor is excusable because of extenuatingcircumstances.When states retaliate,their targets and third parties will better understandthe action as retaliationratherthan unilateral noncooperationor error.Underthe WTO, for example, retaliationmust be centrally authorized,making misinterpretation highly unlikely.
WITHUNCERTAINTY INCREASES ABOUTTHE STATE Conjecture C2: CENTRALIZATION OF THE WORLD.
When states are uncertainabout the state of the world, all may benefit fromjoint efforts to gather and pool information.Scientific activity in Antarcticais coordieconomic organizationshave substantialresearchcapacities nated,and international so that states can sharethe costs of collecting necessary information.In other cases states benefit from collective informationsharingbut have individualreasons not to sharefully or honestly. James Morrowbuilds on the "cheaptalk"literatureto show how regimes can structurecommunicationamong actors to promotemore efficient informationsharingin such circumstances.70
WITHNUMBER. CONJECTURE INCREASES C3: CENTRALIZATION
As numbersincrease, centralizedbargainingreduces transactioncosts by replacing a large number of bilateral negotiations-or even a cumbersomemultilateral negotiation-with an organizational structurethat reduces the costs of decision Centralization also allows states to coordinatetheir operationalefforts to making.71 achieve economies of scale and to ensure that they do not duplicateor work against
68. Axelrod and Keohane 1986. 69. Milgrom, North, and Weingast 1990. 70. See Morrow 1994c; and Farrell and Gibbons 1989. The parallel relationshipthat centralization increases to resolve uncertaintyabout other states' preferencesor types is also likely to hold. The very willingness to allow centralizedinspection by an organizationlike the IAEA containsuseful information about a state's goals even before it generates any informationabout its behavior. 71. See Keohane 1984; and Martin 1992a.
each other. NATO, for example, provides these advantagesthrough a centralized command structurethat allocates tasks.72 Centralizationof informationis also increasingly valuable with larger numbers. RandallCalvertshows how with increasinggroup size the shadow of the futuremay Multilateralcommunicationallows states not be sufficientto supportcooperation.73 to achieve decentralizedcooperationthroughan equilibriumwhere noncooperation is punished by all other states, not just the one that was directly harmed.Because communicationis costly, however, this can be substantiallyimprovedby a centralserves as an informationclearinghouse.Indeed, where a "director" ized arrangement the director can even be viewed as "a third-partyenforcer ... [who] in effect pronouncesa sentence on the deviantplayer, a sentence that will then be carriedout by rationalplayers."74 The InternationalCoffee Organization plays exactly this role in aggregating Moreover, reportsby importingcountrieson coffee shipmentsby exportingstates.75 because decentralizedcooperation typically entails multiple equilibria, centralization is useful in coordinatingbehavior on an agreeable equilibrium.An important example is standardsetting, where intergovernmentalorganizations (such as the InternationalTelecommunicationsUnion) and private organizations (such as the International Accounting StandardsCommittee)provide valuable centralizedcoordination.76 Finally, althoughwe are focusing on main effects, thereis an interactionbetween independent variables that supports conjectures C1 and C3. While decentralized it cooperationis theoreticallypossible with large numbers,77 becomes much more tenuous when even small levels of uncertaintyare introduced.JonathonBendor and Dilip Mookherjee show how centralizationincreases cooperationunder such conditions. In their model a central headquartersis effective because it monitors behaviorand excludes shirkersfrom subsequentbenefitsof the institutionalarrangement.78Such a centralized arrangementcan supporthigher levels of cooperation than can be supportedin any decentralizedarrangement.
Conjecture C4:
MENT PROBLEM. CENTRALIZATION INCREASES WITH THE SEVERITY OF THE ENFORCE-
In the previous conjectures, centralizationalleviates cooperationproblems created or aggravatedby uncertaintyand numbers. But enforcement problems also
72. Abbott and Snidal 1998. 73. Calvert 1995. 74. Ibid., 70. 75. See Bates 1997; and Koremenos 1999a. 76. See Genschel 1997; and Abbott and Snidal 2001. 77. Fudenbergand Maskin 1986. 78. Bendor and Mookherjee 1987 and 1997. Bendor and Mookheree offer a differentiatedview of centralizationand show how a combination(federalism) of centralizedand decentralizedarrangements is most effective for the problem they are examining. Ostromprovides evidence of how small levels of centralizationcan promote otherwise decentralizedcooperation.Ostrom 1990.
790
occur with good informationand small numbers.When the payoff from unilateral defection is significantly greater than from mutual cooperation, concern for the future may not guarantee reciprocity-based,self-enforcing cooperation. In such contexts states may find it optimal to delegate power to a thirdparty to adjudicate and enforce mutually beneficial agreements.79 Concernfor sovereignty,of course, limits the extent to which states will delegate strong coercive capacities to internationalorganizations.But the ability of organizations like the World Bank to withhold resources gives them significantleverage over weaker states. And the informationalcapacities of international organizations to expose states' behavior can influence the activities of even the most powerful statesby imposing international costs or, sometimes, domestic audience reputational costs. Thus states typically obey the findings of WTO dispute-settlementproceedings even though the WTO has no enforcementcapacity. Such mechanismsfall far shortof coercive enforcement,but they can be valuable in "toppingoff' the strictly decentralizedincentives that supportcooperation. Expanding on Bendor and Mookherjee, Edward Schwartz and Michael Tomz show how centralized arrangementshave significant advantages if the central authorityhas the ability to expel shirkersfrom the group.High levels of monitoring will encourage contributionsfrom all actors because shirkersare too likely to be detected and expelled and the value of remainingin the group will increase.80 Even centralized institutions that have no enforcement or even adjudicative capacities may be effective in resolving enforcementproblems.Eric Posner shows thateven if courtsare "radicallyincompetent" determiningfault-that is, they can in determineonly whethera legal agreementexisted but cannot verify whetheractors obeyed it-formalized agreements can create reputationalincentives that enable parties to solve commitmentproblems.81The reason is that the incentive for each costs of the breakdownof the partyto cheat is reducedby the increasedreputational agreementregardless of who is at fault. In a similar vein Lisa Martin shows that internationalorganizations are instrumentalin maintaining support for sanctions partlybecause states do not want to underminethe other benefits providedthrough these organizations.82 Finally, modest internationalcentralization is sometimes effective because it harnessesdomestic enforcementcapacities.The 1998 OECD Anti-BriberyConvention relies on domestic legislation for implementation and on domestic court systems for enforcement, but a centralized inspection system ensures that states
79. Using similarlogic, Lake arguesthat "theprobabilitythat the partnerwill engage in opportunistic behavior decreases with relationalhierarchy."Lake 1996, 14. In other words, as the expected costs of opportunismincrease, hierarchywill be the preferredgovernance structure. 80. Schwartzand Tomz show that the value of centralizationdoes not always increase monotonically with the capacity of the centralagent. Schwartzand Tomz 1997. In their model, an intermediatelevel of monitoringmeans that some shirking will occur so that less talented actors are detected and excluded from the group. 81. Posner 1999. 82. Martin 1992b.
police their own firms. This reinforcesthe point that centralizationdoes not require international agents to have an independentcoercive capacityto effectively promote cooperation. Despite the advantagesof centralizationcapturedin the conjectures,states retain deep-seated concerns, intensified by their risk aversion, about how international institutions might behave. Will resources be squanderedin bureaucraticexcess? Even more important,will internationalagencies expand their authorityover time? Consequently,states view centralizationwarily, and its overall baseline level may remain quite low. Our conjecturesonly express conditions under which states will increase (or decrease) centralizationin response to their environment.For the same reasons, states also are concerned about maintainingtight control over the instituas tional arrangements, indicated in the next set of conjectures. ConjecturesAbout Control Two conjecturesare relevant to the rules chosen to govern institutions: Conjecture VI: Conjecture V2:
AS DECREASES NUMBERINCREASES. CONTROL INDIVIDUAL AMONG ASYMMETRYOF CONTROLINCREASESWITH ASYMMETRY
CONTRIBUTORS (NUMBER).
The first conjectureseems obvious: as the numberof actors increases, the control of any one actor or subgroup of actors decreases.83For example, as the EU has expanded, the leverage of individual members has steadily decreased.84This is because when the numberof actors is large, states must sacrifice individualcontrol to achieve collective benefits.Each state may be adverselyaffected on occasion, and without the veto a state has no unilateral protection-although its ability to withdrawfrom the institutionultimatelylimits its vulnerability.States agree to such a scheme because they benefit from others' inability to veto and strategicallyblock group decisions. An important example is the EU's move toward "qualified majority"voting as membershiphas expanded.85 This conjecturefollows directly from the social choice literatureon voting rules. Brian Barry,for example, shows that for issues that are recurrentand symmetricin
83. Numberhere refers to membersof the institutionwho are eligible to have a say in its operations. This is a good example of our earlier observation that a prior institutionaldecision may be treated as exogenous in considering the adoptionof other rules. Alternatively,membershipand control rules may be determinedtogethersuch that, for example, a decision to have a large membershipis compatiblewith one set of control rules, and a decision to have a small membershipis compatible with another set of control rules. 84. Hosli 1993. 85. A more sophisticatedanalysis would also consider the policy preferencesof governments.Garrett and Tsebelis show how this leads to a considerationof a broaderset of control institutions(for example, the Commission and the Council of Ministers) and to rules regardingother forms of control, such as agenda setting. Garrettand Tsebelis 1996.
792
several senses, majorityvoting maximizes expected utility.86Similarly, the conjecture is supportedby analogy to the theory of the core and noncooperativesolution concepts, where increased power to subgroups (such as through vetoes) leads to paralysisby eliminating mutually agreeable outcomes.87 The second conjecturefollows from an intuition that an actor's control over an institutionrelates to the actor's importanceto the institution.This correspondsto cooperative game-theoretic solution concepts such as the Shapley value, which relates what an actor (potentially) brings to different coalitions to the pay-off the actor receives. When some states contributemore to an institution than othersperhapsbecause they pay more dues or their behavior is vital to the institution's success-they will demand more sway over the institution.Other states will grant this control to ensure their participation-as the UN did to the permanentmembers of the Security Council, whose military and financial support was considered essential to the enforcementof resolutions.88 Membershipand voting rules typically formalize this control in some way, as is the case in the UN Security Council and in the weighted voting in the IMF.
Conjecture V3: INDIVIDUALCONTROL(TO BLOCK UNDESIRABLEOUTCOMES)INABOUTTHE STATEOF THE WORLD. CREASESWITHUNCERTAINTY
Because states are risk-averse, they design institutions that protect them from unforeseen circumstances.Veto power is a standarddesign feature that provides such protection,either to individual states or, in the case of super-majority requirements, to groups of states. A parallel in U.S. politics is the institutionalnorm of universalism,where legislatorsplace a projectin every member'sdistrictratherthan The "theoretrisk being excluded from a (minimumwinning) majorityprogram.89 ical engine" behind the universalistic result is uncertainty and legislators' risk aversion.90 Conjectures C2 and V3 illustrate quite different institutionalresponses to the For problemof uncertainty. example, centralizationof informationcan be increased to remedy uncertainty about the state of the world, with the level of control unaffected.Or super-majority voting may mitigateuncertaintyaboutthe state of the world without changing the level of centralization.In short, control and centralization can be varied independentlyor together to deal with uncertainty.
86. Barry 1979. See also the Rae-Taylortheoremin Rae 1969; and Taylor 1969. Mueller provides an rules. Mueller 1989. Buchanan excellent overview of the issues and a comparisonof majority/unanimity and Tullock argue for the virtues of unanimity in promoting efficient outcomes when there are no transactioncosts. Buchananand Tullock 1962. As decision-makingcosts increase-including the costs of preference revelation (which correspondsto uncertaintyabout preferences)-the case for smaller majoritiesgrows. 87. Shubik 1982. 88. Winter 1996. 89. Weingast 1979. 90. Collie 1988.
Other institutional arrangementsprovide different forms of protection against uncertainty. Escape clauses in effect allow a state to "veto" some institutional dictates only for themselves. Withdrawalclauses allow the more dramaticstep of leaving an institution entirely to avoid undesired outcomes. Such control features blend into what we call flexibility.91 ConjecturesAbout Flexibility Uncertainty about the current or future state of the world presents states with a dilemma. Becoming locked into an institution may lead to unanticipatedcosts or adverse distributional consequences.But by not makinga bargain,states might pass up significant benefits from cooperation. If uncertainty high and anticipatedbenefitsare low, risk-aversestates will avoid is committingthemselves to rigid institutions.But what if the uncertaintyis lower and the potential benefits are higher? Under these more benign conditions, institutional flexibility becomes important. The possibility of adjusting the agreement when adverse shocks occur allows states to gain from cooperation without tying themthat may become undesirableas conditions change.92 selves to an arrangement Conjecture Fl:
WORLD. WITHUNCERTAINTY ABOUTTHESTATEOF THE INCREASES FLEXIBILITY
implicationsof particSimilarly, states may be uncertainabout the distributional ular aspects of an agreement. Koremenos develops a model where states plan to renegotiate all or part of an agreement once they have learned from experience which states benefit the most.93The desirability of renegotiation(versus a single, longer agreement) increases with uncertaintyabout the distributionof gains and decreases with the degree of "noise" in the environmentfrom which the effects of the agreementmust be distinguished.An example is the AntarcticTreaty.Although it has no expirationdate, the treaty was designed to allow states to learn from their experience and modify the agreement over time. One procedurefor modification operatedduring the first thirty years, anotherduring the subsequentperiod. In the first learningphase, the partiesmet biannuallyfor consultations,and the agreement could be changed only by unanimousconsent. Some changes and extensions were to made, such as the follow-on arrangement ban resource extraction.Now that the initial periodhas ended, individualstates can press for renegotiation,this time under
91. We proposedbut laterdroppedthe relatedconjecturethat "individualcontrol(to block undesirable outcomes) increases with the severity of the distributional problem"because it was logically equivalent to conjectureV3. The impact of distributionflowed fundamentallyfrom uncertaintyabout the distribution ratherthan from knowndistributional consequences, which could be dealt with in other institutional ways. The deleted conjecturewas stronglysupportedin the empiricalstudies, so droppingit does not bias the results in our favor. 92. Downs and Rocke 1995. 93. Koremenos 2001.
794
majorityrule. They do so with more certaintyabouthow the agreementoperatesand a better understanding its costs and benefits.94 of Flexibility need not be so formalized. For example, "soft" internationallaw allows states to respond to uncertaintyby designing arrangementsthat are less formalizedthan full legalization. Although often seen as a "failure"of international law, soft law may represent a superior institutional adaptation because of its flexibility.95 Even when states face no uncertaintyaboutproposedagreements,flexibility may resolve distributionalproblems: Conjecture F2:
PROBLEM. FLEXIBILITY INCREASES WITH THE SEVERITYOF THE DISTRIBUTION
Fearon argues that when states lengthen the shadow of the future to solve enforcementproblems, distributionalconcerns become increasingly severe. States Koremenos bargainharderbecause the resultswill affect them for a longer period.96 that in this case states may reduce distributional and bargaining suggests problems, Busch and Koremenos costs, by adopting a more flexible agreement structure.97 show thatundercertainconditions, a series of shorteragreementsstill embodies the shadow of the futurerequiredfor enforcementwhile avoiding the bargainingcosts associated with a single, long agreementin Fearon's model.98 Flexibility has a downside. Renegotiationof treatyterms, as well as dealing with unilateral invocations of flexibility such as escape clauses, is costly. Moreover, individual states have incentives to free ride on an agreement by developing of self-serving interpretations escape clauses that are broaderthan intended. And an opportunity states to "holdup"the cooperativebargain for renegotiationprovides in an effort to increase their own share. Such incentives become greater as more states are partyto an agreement-for the familiarreasons associatedwith collective action.99 Even without these strategic considerations, as more states become involved, modificationbecomes more difficult and time consuming. This reasoning leads to our final conjecture. Conjecture F3:
FLEXIBILITY WITHNUMBER. DECREASES
All else equal, states will introduce less flexibility into institutions with larger numbersbecause largernumbersincrease the costs associated with flexibility more than they increase its benefits. For example, where flexibility takes the form of
94. 95. 96. 97. 98. 99.
This kind of flexibility also solved importantdistributionalissues, the subject of conjectureF2. Abbott and Snidal 2000. Fearon 1998. Koremenos 2001. Busch and Koremenos 2001b. Hardin 1992.
periodic renegotiationof the agreement,largernumberswill increase the associated bargainingcosts. Koremenos shows formally that as renegotiationcosts increase, rational parties to an agreement will renegotiate less often or not at all.100Thus commodity agreementsinvolving forty or so countriesare renegotiatedsignificantly less often than are monetaryagreementsinvolving the G-7. As renegotiationcosts rise, other forms of flexibility become relativelyless expensive. For example, states may switch to more centralized forms of flexibility, such as escape clauses combined with a centralized monitoring institution to keep the moral hazard problem in check or the creation of a quasi-legislative institution empowered to Such changes are consistentwith conjectureC3, adjustthe termsof an agreement.101 that centralizationincreases with number,which brings up the question of design interactions. Finally, note that for some types of flexibility, such as withdrawal clauses, the effects of numberon the form or incidence of the provisions may be minimal. Design Interactions Our simple researchdesign has considerableadvantages,but it also has limitations. Because our definitionsare broad,they encompass significantinstitutionalvariation. The best example is centralization,which includes everything from rudimentary forums for bargaining,throughinformationand monitoringfunctions,to centralized adjudicationand enforcement.Such general conceptions are essential for assessing similaritiesacross cases, but finer conceptualdistinctions are needed to understand the more detailed workings and differences among institutions. The volume's contributorsbegin to do precisely that in the empirical studies that follow. Our bivariaterelationshipscannot capturemore complex interactionsamong the variables.For example, while both large numbersand increaseduncertainty promote centralization,the interactionof their effects may be most significant of all. The most interesting complexities are those that (may) arise because the dependent variables interact among themselves-as "substitutes,""complements,"or "conflicts." Institutionalfeatures may substitutefor one anotherby offering alternative ways to solve a particularproblem. Escape clauses, for example, introduce flexibility to allow hard-pressedstates to avoid the full burdenof their treatyobligations on a decentralizedbasis. An alternative arrangementwould be to require states facing special difficultiesto seek relief from a centralizedinstitutionthat can decide how rules apply to new situations. Thus institutional design can enable choice among different means toward the same ends-that is, a choice among multiple institutionalequilibria. Design features may also complement one another. Membership rules, for example, provide one means to deal with enforcementproblems (conjectureMl),
100. Koremenos 1999a. 101. For a theoretical analysis with correspondingempirical support,see Koremenos 2000.
796
when incentives to defect are especially but these can be enhancedby centralization large. Centralizationmay work either directly as a separatesource of enforcement capacity (conjecture C4) or interactively in making the membership mechanism more effective by providing informationon members' performance.102 Design principles may conflict with one another. Consider an issue with both distributionand enforcement problems. When enforcement is problematic,membership needs to be restricted (conjecture Ml), but when there are distributional problems, it needs to be more inclusive (conjectureM3). Obviously, membership rules cannotremedyboth problemssimultaneously.The only way to circumventthis conflict is to move to a more complex design (such as addressingthe enforcement problem with membership rules and the distribution problem by increasing scope).'03 Our bivariateanalysis cannot fully capturesuch complex interactions.'04 in Finally, our analysis looks at individual institutionalarrangements isolation. Substitutabilities,complementarities,and conflicts arise not only in the design of individual institutions but also in relationships among them. Just as individual featuresof institutionscan complementeach other, so too can differentinstitutions. One way is by vertical nesting, where institutionsthat deal with one issue or region are situatedwithin a largerglobal institution.Vinod Aggarwalhas analyzedexactly this kind of relationship between GATT and various textile arrangements.'05 Likewise, the policymakerswho plannedNAFTA made sure it conformedto GATT trading rules, an issue that will remain importantas both NAFTA and the WTO evolve. We have embracedthese challengesby asking the authorsof the empiricalstudies to begin from our concepts and conjectures.We also asked them to be criticalof the concepts and on the lookout for ways to refine and improve the conjectures.The ultimate value of our conjectureslies less with their individual veracity than with whetherthey spur our collective effort to systematize and refine our knowledge of institutionaldesign. Roadmap to the Rational Design Project The wide range of conjectures (summarizedin Table 1) represents our effort to institutionsfrom a rationalistperspective.The understand design of international the ultimatevalue of our frameworkdependson its ability to explain phenomenaacross a range of substantive issues. The articles that follow take up this challenge by
102. The choice among alternativesmay also dependon interactionswith otherindependentvariables. Thus, the WTO's move toward more centralizeddispute resolution was related to the large numberof states involved. 103. This problem has been central to the analysis of macroeconomic policy in open economies, especially the relationshipbetween the numberof policy goals and the number of policy instruments. Mundell 1962. 104. This problem would bias the empirical results against our bivariateconjectures. 105. Aggarwal 1985.
TABLE1. Summaryof Rational Design conjectures

Ml: M2: increases with the severity of the ENFORCEMENT Restrictive MEMBERSHIP problem PREFERENCES ABOUT increases with UNCERTAINTY Restrictive MEMBERSHIP increases with the severity of the DISTRIBUTION MEMBERSHIP problem increases with NUMBER SCOPE increases with the severity of the DISTRIBUTION SCOPE problem increases with the severity of the ENFORCEMENT SCOPE problem BEHAVIOR ABOUT increases with UNCERTAINTY CENTRALIZATION OF THESTATE THEWORLD ABOUT increases with UNCERTAINTY CENTRALIZATION increases with NUMBER CENTRALIZATION CENTRALIZATIONincreases with the severity of the ENFORCEMENTproblem decreases with NUMBER CONTROL increases with asymmetry of contributors (NUMBER) Asymmetry of CONTROL OF THESTATE THEWORLD ABOUT increases with UNCERTAINTY CONTROL THESTATE THEWORLD OF ABOUT FLEXIBILITY increases with UNCERTAINTY increases with the severity of the DISTRIBUTION FLEXIBILITY problem FLEXIBILITYdecreases with NUMBER
M3: S1: S2: S3:

C1: C2: C3: C4: VI: V2: V3: Fl: F2:
F3:
evaluating our conjectures in the context of many different areas of international politics. The empirical articles all share our rationalistapproach,taken broadly,but they vary widely in other respects. The institutionsexamined cover the full spectrumof internationalpolitics, from environmentalprotection to national security. Some institutions are highly articulatedorganizations;others are much more informal The cases exhibit considerablevariationin key institutionaldimenarrangements. sions, such as centralizationof informationor breadthof membership. We have deliberatelyincluded methodologicaldiversity. Case studies and quantitative approachesare represented.Some analysts develop our conjecturesfurther by using a formaldeductiveapproachto explain the design of institutionsthat affect specific issues; others use a more inductive and empiricalapproachto evaluate and extend the theoreticalframework.While most of the studies treat states or international organizations as their central actors, others focus on private international actors, such as firms and private courts, or relax the unitary actor assumption to incorporatekey domestic political factors. Most of the studies treat institutional rational design as a deliberaterationalchoice; one, however, focuses on "indirect" design driven by actors' selection among available institutionalalternatives.The firstthree articlesdevelop the theory in specific contexts and enrichit by connecting it to specific empirical cases. The next five articles use the theory as the basis for intensive empirical analysis of a specific issue-area. Andrew Kydd looks at NATO enlargement and investigates the causes and consequences of NATO's membershipcriteria.NATO enlargementhas built trust among the potential entrants but weakened it between NATO and Russia. The membership criteria are fairly restrictive: new members must have firmly entrenched democracies, civilian control of the military, and no ethnic or border
disputes with their neighbors. These restrictive criteria build trust among new membersby diminishinguncertaintyabout their preferences;they also mitigate the distrust generatedin Russia, by showing that NATO is not just expanding willynilly to include any state that wants to join. Peter Rosendorff and Helen Milner look at one of the most common and controversial features of trade agreements: escape clauses. This design feature allows states to enter into agreementsthey might not otherwise accept because of unforeseeable contingencies. But escape clauses must be costly, or else countries might use them cynically to abandon agreements that are merely inconvenient. Rosendorffand Milner develop a formalmodel that shows how states design escape clauses to balance these considerationsand facilitate agreement. RobertPahre asks why states often "cluster"negotiationswith multiple states at the same time. He develops a model of clustering, which he tests on nineteenththe century trade relations. But his analysis is equally insightful for understanding use of negotiating rounds in the postwar GATT/WTO.Clusteringoccurs in other issue areas as well. It is especially important when states are committed to most-favored-nationpolicies because these exacerbate distributionalproblems by linking every bilateral trade negotiation to every other negotiation. Clustering is importantbecause it helps states resolve these distributionalproblems. Ronald Mitchell and PatriciaKeilbachuse their study of environmentalissues to investigate institutionaldesign when asymmetricrelationshipsexist among actors. Sometimes "upstream"states create pollution, and "downstream"states are its victims. Pollutershave no incentive to join an institutionto reducepollutantsunless the institution's scope includes issues they might benefit from. Asymmetry occurs in anotherway as well. Polluting states can be strongeror weaker than the victims. Mitchell and Keilbach show that weak victims seek institutional designs with positive linkages or rewards, whereas strong victims prefer negative linkage or sanctions. Walter Mattli highlights the growth of private institutions to arbitrateinternational business disputes. Private tribunalsare often faster, more discreet, and less expensive than public courts. They can be designed to focus closely on specific commercial practices within an industry,a kind of expertise courts rarely possess. The demandfor arbitration been so strong that business groups have produced has a multitude of arbitrationtribunals. The strengths and weaknesses of different designs lead business partnersto select a tribunal to handle disputes as part of commercialcontracts.Theirchoice, Mattliargues,dependson the numberof parties involved and their uncertaintyabout the future state of the world and each other's behavior. Thomas Oatley deals with a very public institution, the system of multilateral trade and payments for Europe's postwar reconstruction.Two major design problems faced Europeans. One was distributional:who would bear the costs of to reserves.The United adjustment tradeimbalances?The second was hard-currency States was willing to provide dollars throughthe MarshallPlan but feared it might lead to bloated debts ratherthan disciplined development. Oatley shows how the
payments union begun in 1950 resolved these issues with a series of interrelated design features:centralized trade and credit balances, flexible administration,and relatively weak enforcement. When fighting breaks out, enemy soldiers are frequently seized as prisoners of war. States have joint treaties to ensure that prisoners are treated humanely and JamesMorrownotes modify them to cope with new types of war and imprisonment. that a workable treaty design must affect the behavior of front-line troops who treatiesare designed with thatin mind. actuallycaptureprisoners;twentieth-century Moreover, because these treaties entail some costs, ratifying them sends signals about national intentions. Standardsfor treatmentare generally straightforward, partly to make them easily understoodby soldiers, partlyto resolve any wrangling over the distributionof burdens. John Richards deals with the institutionaldesign of the global aviation regime. States had to decide whether markets or regulation would govern air routes and fares. Their choice of regulationwas promptedby nationalsecurityconcerns,which were closely tied to aeronauticsand to states' desire to promote high-technology industriesat home. Once on the regulatorypath, states faced the complicatedtask of building effective internationalinstitutions. Richards shows how the regulatory institutionsthat emerged were profoundly shaped by the particularfeatures of the industry,including the large numberof states involved and their uncertaintyabout one another's behavior and future conditions. The volume concludes with two articles. We invited Alexander Wendt to comment on the projectfrom an "external" perspective.Wendt is both sympathetic to our enterpriseand skeptical of it. He questions our decision to focus on rational choice explanationswithout directly engaging either competing approachesor what but he believes are complementary "deeper" explanations.Wendtfurtherarguesthat our analysis is insufficiently "forwardlooking" to address importantnormative concerns. While we do not fully agree with Wendt's critique, his article provides insight for both insiders and outsiders about the limitations of our approach. In the final article we summarize the findings. We also combine internal and external critiques of what the volume has accomplished and consider how our rationalistapproachcan be improvedby addressingquestions raised by alternative perspectives.

International Organization Foundation

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

International Organization Foundation

Uploaded by

Copyright:

Available Formats

International Organization Foundation

The Rational Design of International Institutions

The Rational Design of International Institutions 763

Scope of issues covered (SCOPE)

The Rational Design of International Institutions 765

The Rational Design of International Institutions 767

768 International Organization

The Rational Design of International Institutions 769

are not the only

770 International Organization

The Rational Design of International Institutions 771

772 International Organization

The Rational Design of International Institutions 773

The Rational Design of International Institutions 775

game-theoretic logic allows us to connect them to the dependent variables of institutionaldesign.32

The Rational Design of International Institutions 777

36. Snidal 1994.

778 International Organization

The Rational Design of International Institutions 779

Conjectures About Rational Design

The Rational Design of International Institutions 781

3. Transactioncosts: Establishingand participatingin internationalinstitutions

Institutions 783 The Rational Design of International

784 International Organization

The Rational Design of International Institutions 785

The Rational Design of International Institutions 787

788 International Organization

Institutions 789 The Rational Design of International

Institutions 791 The Rational Design of International

The Rational Design of International Institutions 793

94. 95. 96. 97. 98. 99.

The Rational Design of International Institutions 795

Institutions 797 The Rational Design of International

TABLE1. Summaryof Rational Design conjectures

M3: S1: S2: S3:

798 International Organization

The Rational Design of International Institutions 799

You might also like